Modification of anaerobic digestion model No.1 with Machine learning models towards applicable and accurate simulation of biomass anaerobic digestion

This work proposed a so-called M-ADM1 model for anaerobic digestion simulation, which uses machine learning model to predict the kinetic parameters in anaerobic digestion model No.1 (ADM1). A total of 75 biomass samples were used to establish the machine learning model. Inputs used to predict the kinetic param-eters included C, H, O, N, S contents, and digestion temperature. The sensitivities of 17 kinetic parameters were evaluated, and 7 kinetic parameters with the highest sensitivities were selected as model outputs. After model optimization, the average R2 for predicting the 7 kinetic parameters reached 0.92, and the root mean square error reached 0.167. The accuracy of the overall M-ADM1 expressed by Theil inequality coefficient of municipal solid waste, kitchen waste, and sludge were 0.0163, 0.0327, and 0.0361, respectively. The results validated the hypothesis that accurately predicting some crucial intermediate parameters using machine learning models could enhance the performance of tradition ADM1.